A Plugin Pack for Cursor, Built by the People Who Make Cursor
If you've spent any time wiring AI coding agents into your workflow, you've probably hit the same wall: the agent is capable, but it doesn't know how you work. It doesn't know your CI setup, your review standards, or the conventions your team actually follows. You end up re-explaining the same context every session. The cursor/plugins repository is Cursor's attempt to fix that—a collection of official plugins that package up reusable agent behaviors for specific tools, frameworks, and workflows.
What It Does
This repo is a monorepo of Cursor plugins. Each plugin lives in its own standalone directory at the repository root and ships with a .cursor-plugin/plugin.json manifest. That structure keeps things simple: one folder per plugin, self-contained, easy to browse or borrow from.
The plugins themselves span a fairly wide range. Some are productivity integrations—Gmail, Google Drive, Google Calendar, and Google Docs plugins that let you search, read, draft, and manage email or files, schedule meetings, and so on. Others are developer-tooling oriented: create-plugin scaffolds and validates new agent plugins, cursor-sdk covers building apps and automations with the TypeScript SDK, and agent-compatibility runs CLI-backed compatibility scans that audit startup, validation, and docs against reality. There's also a growing set of workflow plugins contributed by Cursor staff and outside developers, aimed at things like code review, orchestration, and memory.
Why It's Cool
The manifests make it a reference, not just a download. Because every plugin has its own plugin.json and lives in a clean root-level directory, you can read any of them as a working example of how to structure your own plugin. That's a quietly useful property—most plugin ecosystems bury their conventions in docs rather than shipping them as browsable code.
The workflow plugins are opinionated in interesting ways. pstack, by Lauren Tan, is described as helping you "write less, but higher quality code" through rigorous agent workflows you can parallelize. dyl-stack layers Dylan Gattey's personal agent style on top of pstack—root causes over symptom patches, terse verified delivery, a PR review that fits in a paste. These aren't generic helpers; they're specific methodologies encoded as plugins, which is a different and more honest approach than pretending there's one right way to use an agent.
Some of them tackle genuinely hard coordination problems. orchestrate fans large tasks out across parallel cloud agents with planners, workers, verifiers, and structured handoffs. thermos runs what it calls "thermo-nuclear branch review"—deep security and correctness audits, harsh code-quality rubrics, parallel subagents, and optional merge-ready PR flows. advisor is a simple but smart idea: consult a stronger model before major decisions, when you're stuck, and before declaring done.
There's attention to the unglamorous stuff. cli-for-agent documents patterns for designing CLIs that coding agents can run reliably—flags, help with examples, pipelines, errors, idempotency, dry-run. continual-learning handles incremental transcript-driven memory updates for AGENTS.md, restricted to high-signal bullet points only. These are the details that determine whether agents actually work in practice, and it's good to see them addressed directly rather than glossed over.
How to Try It
- Head to the repository at github.com/cursor/plugins.
- Browse the plugin table in the README to find something relevant to your workflow—the table lists each plugin's name, author, category, and marketplace description.
- Open the plugin's directory to inspect its
.cursor-plugin/plugin.jsonmanifest and any associated files. - If you're building your own, start with
create-pluginto scaffold and validate a new plugin rather than hand-rolling the structure.
Since this is a plugin collection rather than a library, there's no install command to run—you're working with the plugin directories directly. The README is the source of truth for what's currently available, and the table format makes it easy to scan by category (Utilities, Developer Tools, Productivity).
Final Thoughts
This is best suited to developers already working inside Cursor who want to stop re-teaching their agent the same things, and to anyone curious about how agent plugins are structured in practice. The mix of official and community-contributed plugins is a good sign—it suggests the format is flexible enough to encode personal workflows, not just vendor integrations. If you've been meaning to build a plugin of your own, reading a few of these manifests is probably the fastest way to understand the shape of the thing.
Follow @githubprojects for more developer tools and open source projects.